Artist’s Roundtable - The artists’ take on Generative AI
Image: Artists Eryk Salvaggio, Lex Fefegha, Caroline Sinders, Amelia Winger-Bearskin
When looking for actionable strategies, insights and recommendations, often we look to researchers, consultants and external experts. It’s crucial though, that we gain insight from those at the coal-face; and with respect to Generative AI and the creative industry, that’s the artists who are making the work and building the tools.
The New Real’s creative agent, Caroline Sinders, sat down with three other artists working with Generative AI – Lex Fefegha, Eryk Salvaggio and Amelia Winger-Bearskin – to explore future landscapes for Generative AI Arts, find out what co-creation between AI and artists can look like, and - simply - capture what artists want from both AI and those influential people in the broader ecosystem of funding, curation, museums and policy-making.
What follows is an edited-for-clarity-and-brevity eavesdrop into their conversation.
—
Caroline Sinders: So tell me why you all work with Generative AI and what are your thoughts about it as a tool?
Lex Fefegha: I've always been interested in building software, and creating stuff for creators to create interesting stuff, but I knew one thing about AI: if we're training this on the same data sets that currently exist, similar to if you were to put that AI tool in a criminal justice situation, the dataset is going to have a high chance of bias. A lot of the work I've been doing is to help tech companies make sense of Generative AI in terms of what feature they should build and what role Generative AI should actually play in society. I recently worked with IBM Watson; I was asked to create an interactive installation with Generative AI, which would look at different moments in history. The context and concept behind this interactive installation was the weather as the original influencer in history: we took moments in history and said, what happened if the weather was different that day? I learned a lot as this was the first time I'd ever worked with Mid Journey (where before I had trained my own models from scratch), and so I had to learn how to be a good prompt engineer. You don't have the control you might want to have; you have to learn how the machine understands a prompt or how it sees an image, especially where I was talking about hip hop and blackness and things that are not necessarily ‘default’. There’s a lot of bias in these things, especially when it comes to prompt engineering.
Eryk Salvaggio: I first started getting involved in Generative AI when I realized that it did not produce very accurate images of black women. You got distortions: you got lower resolution, blurrier images and more errors. And thinking through that as opening up the dataset and seeing there's very few images of black women in this dataset. And then realizing that is also being used to train surveillance systems, being used to train all kinds of different processes and realizing that this relationship between the image and the bias that is in the datasets is circulatory. And thinking about that as an artist was interesting, because you could test and prove that by making work. If you're making work with GANs (Generative Adversarial Networks), you're intentionally biasing datasets, and you can twist those biases. Instead of relying on GANs to train images on one category, say: flowers, you could train it on flowers and ballet dancers, and get these hybridizations. This oriented me as a way of thinking about that really strong link between Generative AI and the underlying issues that are present in the ideologies of AI and thinking through how we can sort of twist those.
Amelia Winger-Bearskin: Nowadays, we are up against a lot of assumptions and misinformation when we talk about Generative AI – I often combat this by being very descriptive about what my work, like: “in this work, I'm using an AI painting technique, this is what the algorithm does for that, this is why I chose that, this is the maths behind it.” I feel like I'm combating a lot of language around it being magical, for example. But people are more aware that it's biased, which is good, but then I’m combating them being angry that any artists would choose to engage in an unethical, unregulated, terrible thing that's stealing all of our jobs. In the past it may have sounded alarmist to be worried about the repercussions this thing could have in the future. Well, now we're in that future. For example, what does it mean when AI is trained on the basis of work that has been stolen? People say that there are no laws regulating the use of AI, and yet that's not true: existing basic copyright laws, they're already violating those. I speak a lot to people who are part of the SAG-AFTRA strike – that is the first union to be public about Generative AI and its impact on their industry.
CS: A lot of people see it as inherently contradictory that artists might use a tool that they don't absolutely endorse. I’m somebody who has made an entire art career out of using tools that I'm deliberately focusing on strengthening critique of and understanding from the inside-out. We need people who know these tools to also be on the side of the people who are resisting those tools.
AWB: For those of us who have been in the art and technology space for a long time: we are failing if we are not building bridges to the current pain points there are in an industry that is adjacent to ours; which is also ours – we may make work commercially in different ways than SAG-AFTRA, but at the heart, we're all artists. They're part of our community, and we're failing if we're not showing up by saying: for 20 years we've been studying this, we've been ringing this bell and no one's been listening to us. Now other people are actually on the street; on a picket line. We have to remind them that our field has always been about challenging the ethics of what technology is doing, positively and negatively in our field. And in being that pushback and that check to how far something can infringe upon our human rights both by understanding and knowing it. We know it because we can use it. We know it because we've helped build it.
CS: So much of the history of art is also about technical innovation, like with the onset of photography for example. All of this ends up changing and impacting the ways in which we make art and it feels like sometimes we forget that so much of art is not only a dialogue and critique of technology by doing, that it's also an engagement of the process itself. When you see a seasoned artist use a tool, it's sometimes very different than how the creator of that tool conceptualised it. There's something interesting that artists can do, where art making doesn't stop with the generation of this one thing; it’s the context of how I'm going to use this. Some of the art I've seen that uses Generative AI is often a very big series of work, it’s not a singular image, right? It's a part of something much larger.
AWB: I really love the moment when any tool becomes truly democratised. Like when Photoshop just first came on the scene. I'm so old that I remember the controversy because people found a couple of versions of supermodels where it was very obvious that they had done a bad Photoshop – little did they know that every single one of the images was Photoshopped! – but then people were saying we should outlaw Photoshop, no one should be able to use Photoshop, it's terrible for our society; even though it was already pervasive. But then you started seeing things like Gimp and other Photoshop clones that were free and available online, then you got the good memes – the moment that a child can access this technology, stuff gets interesting. That's where the culture-jamming starts, because they are looking at this as a tool of play, not just a tool of industry to make supermodels look even more perfect on the front cover of Vogue or whatever. We wouldn’t have had memes if it wasn't for these knockoff versions of Photoshop. And we wouldn't have had that until Photoshop became truly democratised and understood by the masses – even though it started off with a panic. We're at that moment of total democratisation around Generative AI where a very young child can type something in and see a response and then start playing with that: “What is SpongeBob made of Dorito tacos that's riding a skateboard?” I mean, these are the things that my students do their first time playing around with these tools. They make very, very funny images. It's very human, rather than very industry. I love this moment for us as artists because we get to see what anyone and everyone would use this tool for. It was hard to be in this space until this moment occurred, because it felt too rarefied – I love that the floodgates have been opened. It challenges all of us to make sure that we're doing work that is actually culturally relevant, that pushes back, that is radical, that has some type of revolution baked into it of the world that we want to see.
ES: It's sort of like it's like paints, right? For a long time, purple was this very expensive colour, and then they found a way to synthesise purple and now everyone has access to purple. Now we don't care if there's purple in your painting; it's what you are doing with the purple. There's a lot of really bummed out AI artists who don't understand why they are making work that they think is really visually compelling, and no one is interested in it. They don't quite realise that if it's democratised for you, it's also democratised for everyone else. Making a compelling visual image is no longer interesting; what you actually have to do is try to think about all of the affordances this technology makes available to you. Throw away the instruction manual, figure out where you can push these systems in directions that the tools are not necessarily designed for, but that give you a kind of a unique angle on what you can do with that.
CS: Something about democratisation that is so accurate, is that it's so human. With technology like Generative AI, it’s always kind of off, or it's too polished. It’s a very similar conversation that I think happened with painting, if you considered photography as pure visual mimicry. What we then have is experimental painting, like the onset of Cubism, right? Something a photo can't do. One of the things I'm wondering now with Generative AI: are we going to see the return of people building physical sets, in even new media, or will we see the insistence on really beautiful installations? Will we see people scanning objects where all the mistakes are in the object, where you see the wood-grain; the things that Generative AI does not do; when everything is perfect, in a weird way?
ES: I work with technology in order to despise it properly, which is one of the two big quotes that we have at the Algorithmic Resistance research group. One of the things I'm interested in is: what do we mean when we say imperfection? We have this idea that what we have democratised is not creativity, but instead simply access to the ability to produce an image that we think passes as creativity. One of the things I've realised is that someone stole the definition of creativity from you: they told you it was about making a perfect image. Creativity is the process of trying and failing to make a perfect picture, and when you take that away, you are not democratising creativity or democratising the art process. What you're doing is you're automating that and you are depriving people of that challenge and the joy of discovering your own limitations and working around them. By automating the production of perfect images, what we're actually doing is skirting the entire idea of what creativity actually is. It’s snobby to say, you didn't really make those AI images, and I actually don't think it’s true. I think the typing prompts can be a way of doing art. But we don't talk about it that way. The focus is so much on the outputs of the system that are being scraped out by all this data processing, and you're steering through it. Like what is your struggle in that creative process? That's actually what makes things interesting.
CS: As artists: what do we want from AI? Are there concrete actual requirements for the AI community? Consent, credit and compensation I think is a major thing for me. I wish we could have nuanced conversations about Generative AI without it sounding alarmist, though I’m cognizant of the fact that I do think that this is going to impact aspects of the creative industry. I wouldn't be surprised if we see smaller and smaller fellowships, or a lot more confusion over the artistic and creative practice. Underpinning the global conversation are the misconceptions of how much creativity is worth and how long it takes to be creative, how long it takes to work on a piece of art. This capitalism hellhole we're in: every hour is subdivided into billable minutes, and with the rise of the gig economy, suddenly there’s an expectation that you’ll create a piece of really good work, that you are underpaid for, really quickly. You need a lot of time to be creative. You just need time to sit down and stare at a screen sometimes or look at something that has nothing to do with your work. That is part of the process. And I worry that this further flattens that, and puts us in a place where we're all urgent all the time.
AWB: I think it is important for us to make space for the unknown and to remind people that we're at the beginning of this journey with AI. Nothing has yet been decided; we haven't yet finalised how these tools are going to be used or should be used. Even though it seems like it's moving so fast each year. People always want to be like, what's the call to action? And I think maybe pushing back and saying it's okay to have ambiguous feelings and conversations and thoughts around this for those who are not immersed and bathing in this strange, murky water that we are all of the time. As we interface with policymakers and fellowship directors or other people in the art sector, we need to remind them that we can keep this ambiguousness longer and that will be beneficial for many of us to not come to immediate conclusions.
LF: I'm currently in a place where I'm trying to work with companies who exist in the business of creativity and understand how Generative AI could help them. I want to understand the relationship with labour and how, from a capitalistic standpoint, that influences the way Generative AI is used. I've always sat more on the design side, rather than the artists: I know, innovation is the weapon of capitalism. So for me it’s about how does this make sense and how, if these technologies are going to be implemented, how can it be done in responsible ways? I'm just in a place of learning and observing and trying to make sure I can keep a roof over my head at the same time. I do notice where I sit though.
ES: I liked that point about connecting labour and the idea of showing the work in these systems because they really are designed for erasing work. They erase work by collecting datasets of artists and putting them together and not attributing them. They erase work in that the image you get sort of appears suddenly. You don't see the system struggling with creative choices. Revealing labour in the process is really interesting. I've been having a lot of conversations about Bunraku which is this Japanese puppetry style, where the puppeteers are visible onstage. Something I read which I liked: seeing the performers holding the puppet strings, how could anyone mistake the puppets for a dog? And I think that's a really cogent summary of where we are now: if we could see the labour that went into these systems, the labour that went into the art that goes into datasets that makes those pictures; if we could reveal that somehow, we would perhaps no longer have the illusion that these systems are gods or magic boxes, and I think that would be really important to do.
